JP7697308B2 - Monitoring device, monitoring program, and monitoring method - Google Patents

Description

The present invention relates to a monitoring device, a monitoring program, and a monitoring method, and can be applied, for example, to a monitoring device that measures the soundness of a structure based on the vibration of the structure.

Traditionally, many attempts have been made to install various sensors on factory equipment and infrastructure structures (e.g., bridges and columnar structures) to measure the soundness of the objects. Many objects have a natural vibration frequency (hereafter referred to as "natural frequency"), and it is known that the natural frequency changes when an abnormality such as deterioration occurs in the structure.

On the other hand, in structures where girders are supported by cables, such as cable-stayed bridges, it is necessary to periodically check whether the tension assumed at the time of design is being applied to each cable. Traditionally, the method used was to directly measure tension using a jack or similar device, but this required a lot of time and effort. For this reason, measurement methods that use vibration have been attracting attention in recent years.

In Non-Patent Document 1, a highly accurate acceleration sensor is attached to a cable to measure the vibration of the cable, which is constantly vibrating due to wind, etc., without applying any excitation to the cable. The obtained vibration waveform is then Fourier transformed to obtain a frequency spectrum.

It is generally known that vibrations propagating through a structure have peaks not only at the natural frequency, but also at frequencies that are integer multiples of the natural frequency (hereafter referred to as "natural frequency groups"). The same is true for cable vibrations, which basically have peaks at the natural frequency and integer multiples of the natural frequency.

It is known that there is a certain relationship between cable tension and natural frequency, as shown in Equation 1 in Non-Patent Document 1, and tension can be estimated from the natural frequency. In general, natural frequencies are low, at a few Hz or less, making it difficult to measure them accurately using a sensor. Furthermore, there is always an error in measuring frequency. Therefore, more accurate measurements can be made by using a group of natural frequencies.

For example, with regard to infrastructure structures, the vibration source is a passing vehicle or wind, so the environment is not one in which constant vibration can be obtained. In such cases, there may be situations where there is no vibration and no peak is obtained, or where the peak changes significantly due to an abnormality on the vibration source side. Due to these problems, it is difficult to determine the natural frequency group using only the spectrum obtained from a single measurement, and it is necessary to use a superposition of spectra obtained from multiple measurements. It is not as simple as simply extracting the natural frequency group.

To address these issues, traditionally, humans have had to visually determine the natural frequency, or normal vibration parameters such as the natural frequency and thresholds have been set in advance. Efforts are also being made to automate these settings using machine learning, etc.

Dai Wakabayashi and 3 others, "Tension Estimation and Soundness Assessment of Cables in Cable-Stayed Bridges Using Higher-Order Vibration Methods", 73rd Annual Academic Conference of the Japan Society of Civil Engineers, August 2018, https://sensor-japan.jp/pdf/document/doc_bridge_h3008n.pdf

As mentioned above, tension can be estimated from the natural frequency, but since natural frequencies are generally low, at a few Hz or less, it is difficult to measure them accurately using a sensor. Also, there is always error in measuring frequency. Therefore, more accurate measurements are expected to be possible by using a group of natural frequencies that have the characteristic that peaks appear at frequencies that are integer multiples of the natural frequency.

For example, with regard to infrastructure structures, the vibration source is a passing vehicle or wind, so the environment is not one in which constant vibration can be obtained. In such cases, there may be situations where there is no vibration and no peak is obtained, or where the peak changes significantly due to an abnormality on the vibration source side. Due to these problems, it is difficult to determine the natural frequency group using only the spectrum obtained from a single measurement, and it is necessary to use a superposition of spectra obtained from multiple measurements. It is not as simple as simply extracting the natural frequency group.

To date, the problem described above has been addressed by having humans visually determine the natural frequency or by setting normal vibration parameters such as the natural frequency and thresholds in advance. Efforts are also being made to automate these settings using machine learning, etc.

However, as the number of structures to be monitored increases, it is not realistic for humans to set the natural frequency of each structure. Even if the settings are automated, the settings still need to be checked, and it is difficult to ensure that they have been set without error. In addition, the appropriate settings may change over time, and it would be extremely time-consuming for humans to set parameters and check them periodically.

In view of the above-mentioned problems, the present invention aims to provide a monitoring device, a monitoring program, and a monitoring method that can recognize the natural frequency of each structure from multiple spectrum data, automatically estimate parameters such as tension using the natural frequency, and automatically determine abnormalities in the structures, without having to set a natural frequency for each of one or more structures.

In order to solve such problems, the first monitoring device of the present invention is characterized by comprising: (1) an analysis means for detecting peaks appearing in a frequency spectrum obtained based on a measurement signal from a sensor provided in a structure, and determining information about the peaks for each peak; (2) a storage means for storing information about each peak for each measurement; (3) a grouping means for grouping the frequencies at which each peak is detected by frequency bands based on information about each peak obtained over multiple measurements; (4) a natural frequency group extraction means for extracting a plurality of frequency bands belonging to a natural frequency group from the plurality of frequency bands grouped by the grouping means; and (5) a natural frequency estimation means for estimating a natural frequency based on the plurality of frequency bands belonging to the extracted natural frequency group.

The second monitoring program of the present invention is characterized in that it causes a computer to function as: (1) an analysis means for detecting peaks appearing in a frequency spectrum obtained based on a measurement signal from a sensor installed in a structure, and determining information about the peaks for each peak; (2) a storage means for storing information about each peak for each measurement; (3) a grouping means for grouping the frequencies at which each peak is detected by frequency bands based on information about each peak obtained over multiple measurements; (4) a natural frequency group extraction means for extracting multiple frequency bands belonging to a natural frequency group from the multiple frequency bands grouped by the grouping means; and (5) a natural frequency estimation means for estimating a natural frequency based on the multiple frequency bands belonging to the extracted natural frequency group.

The third monitoring method of the present invention is characterized in that (1) an analysis means detects peaks appearing in a frequency spectrum obtained based on a measurement signal from a sensor installed in the structure, and obtains information about the peaks for each peak, (2) a storage means stores information about each peak for each measurement, (3) a grouping means groups the frequencies at which each peak is detected by frequency bands based on the information about each peak obtained over multiple measurements, (4) a natural frequency group extraction means extracts multiple frequency bands belonging to a natural frequency group from the multiple frequency bands grouped by the grouping means, and (5) a natural frequency estimation means estimates a natural frequency based on the multiple frequency bands belonging to the extracted natural frequency group.

According to the present invention, the natural frequency of each structure can be recognized from multiple spectral data, and parameters such as tension can be automatically estimated using the natural frequency, allowing automatic abnormality judgment of the structure to be performed.

1 is a diagram showing an internal configuration of a monitoring device according to an embodiment; 1 is an overall configuration diagram showing the overall configuration of a monitoring system according to an embodiment; FIG. 2 is an internal configuration diagram showing the internal configuration of a sensor device according to the embodiment. 4 is a flowchart showing a processing operation in the monitoring device according to the embodiment. FIG. 2 is an explanatory diagram for explaining a method of deriving prominence according to the embodiment; 1 is a flowchart showing a procedure for deriving prominence according to an embodiment. FIG. 4 is a spectrum diagram illustrating an example of a frequency spectrum in the embodiment. FIG. 11 is a diagram illustrating an example of data results measured on consecutive days by the monitoring device of the embodiment. FIG. 9 is a diagram showing an example of peak data belonging to some groups among the data results of FIG. 8 . FIG. 11 is a diagram showing frequencies for each group in an embodiment.

(A) Main Embodiments Hereinafter, embodiments of a monitoring device, a monitoring program, and a monitoring method according to the present invention will be described in detail with reference to the drawings.

(A-1) Configuration of the Embodiment (A-1-1) Overall Configuration FIG. 2 is a diagram showing the overall configuration of a monitoring system according to an embodiment.

In FIG. 2, the monitoring system 1 has a parent device 10, multiple sensor devices 20 (20-1 to 20-n; n is a positive integer) as child devices, and a monitoring device 30.

Sensors of sensor devices 20 are installed in the structure to be monitored, and the sensors measure the vibrations of the structure. The sensor devices 20 transmit information including the measurement signals (measurement values) measured by the sensors to the parent unit 10. The parent unit 10 acquires the measurement signals measured by each of the multiple sensor devices 20 from each of the sensors, and provides the information acquired from each sensor device 20 to the monitoring device 30. The monitoring device 30 collects the sensor measurement signals from each sensor device 20, and uses the collected measurement signals to monitor the state and condition of the monitored object.

There are various types of "structures" to be monitored, but the structures are infrastructure structures, such as bridges, that are not constantly subjected to constant vibrations. Of course, the present invention can also be applied to determining abnormalities in structures that are constantly subjected to constant vibrations. In this embodiment, an example is shown in which the present invention is applied to structures such as cable-stayed bridges and extradosed bridges, whose girders are supported by diagonal cables or members.

The sensor device 20 mainly includes a sensor, a computing device for processing data, communication means, a clock, a timer for waking up the entire device from a sleep state, a battery, etc. The sensor included in the sensor device 20 is an acceleration sensor 203 (see FIG. 3), which is fixedly installed in the structure. The acceleration sensor 203 may also be installed in the structure when measuring the soundness of the structure. The sensor device 20 transmits measurement signals (measurement values) intermittently measured by the acceleration sensor 203 to the parent unit 10.

The parent device 10 is a wireless device that transfers information including measurement signals received from the sensor devices 20 to the monitoring device 30. The parent device 10 is capable of communicating with multiple sensor devices 20, and transfers information received from each sensor device 20 to the monitoring device 30.

The parent device 10 may have the same configuration as the sensor device 20, in addition to the forwarding function. In this case, the parent device 10 may directly transmit the measurement signal measured by its own sensor to the monitoring device 30. As another variation, the parent device 10 may have all or part of the functions of the monitoring device 30 described below.

The monitoring device 30 collects and saves (accumulates) the measurement signals measured by the acceleration sensor 203 from each sensor device 20 via the parent device 10. The monitoring device 30 performs a Fourier transform on the measurement signals of the acceleration sensor 203 to derive a frequency spectrum. The monitoring device 30 analyzes the shape of the frequency spectrum, extracts specific frequency ranges in which peaks appear when the structure is in normal operation, and accumulates peak data for each specific frequency range.

At this time, the monitoring device 30 measures and accumulates peak data for each specific frequency range extracted from the frequency spectrum for each measurement by the acceleration sensor 203. For example, if a training period of several days is set and the acceleration sensor 203 measures once a day, the monitoring device 30 measures and accumulates peak data for each specific frequency range extracted from the frequency spectrum once a day. In this way, by setting up a training period, peak data for each specific frequency range can be accumulated, improving accuracy.

In addition, the monitoring device 30 uses a large amount of peak data in a specific frequency range to group the data, and obtains and accumulates peak data in a specific frequency range that is estimated to be the natural frequency of the structure.

Furthermore, the monitoring device 30 extracts a group of natural frequencies from multiple specific frequency ranges (groups) obtained by multiple measurements, and estimates the natural frequency using the difference value between a certain frequency in the group of natural frequencies and an adjacent frequency. Note that the method of extracting the group of natural frequencies and the method of estimating the natural frequency will be explained in detail in the section on operation.

In FIG. 2, an information communication system is illustrated in which a sensor device 20 as a child device transmits information to a monitoring device 30 via a parent device 10, but this configuration of the information communication system is merely an example. As long as the sensor device 20 can transmit information to the monitoring device 30, the configuration of the information communication system is not limited to the configuration in FIG. 2.

The network including the parent device 10, the sensor device 20, and the monitoring device 30 may use a wireless line or a wired line. For example, the sensor device 20 and the parent device 10 may constitute a sensor network SN. The communication method of the sensor network SN can be a low-speed wireless communication method represented by a specific low-power wireless method. For example, a wireless network standard represented by IEEE802.11a/b/g/n, or a wireless communication method such as IEEE802.15.4 or Bluetooth (registered trademark) may be used. The sensor device 20 and the parent device 10 are each assigned a unique address (e.g., MAC address, short address, IP address, etc.) on the sensor network SN.

The network between the parent device 10 and the monitoring device 30 may be a backbone network NT. The backbone network NT is, for example, the Internet or Ethernet (registered trademark). The backbone network NT may be a wireless line or a wired line.

(A-1-2) Internal Configuration of Monitoring Device FIG. 1 is a diagram showing the internal configuration of a monitoring device 30 according to an embodiment.

In FIG. 1, the monitoring device 30 has a communication unit 301, a control unit 302, a peak data analysis unit 303, a grouping processing unit 304, a data accumulation unit 306, an eigenfrequency group extraction unit 307, and an eigenfrequency estimation unit 308.

The communication unit 301 is a communication interface that communicates with the network connected to the parent unit 10.

The control unit 302 is a device or processing unit that controls various functions of the monitoring device 30. The control unit 302 may be configured, for example, as a device having a CPU, ROM, RAM, EEPROM, an input/output interface, etc. The processing may be realized by the CPU executing a processing program (e.g., a monitoring program, etc.) stored in the ROM.

The control unit 302 has a physical numerical value derivation unit that derives values (physical numerical values) indicating the state of the structure, such as tension, based on the value of the natural frequency estimated by the natural frequency estimation unit 308.

The control unit 302 also has a warning unit that outputs a warning when a value indicating the state of the structure, such as tension, deviates from a preset range or deviates from the range of values obtained in past measurements.

The data storage unit 306 stores information for determining the soundness of the structure. For example, the data storage unit 306 stores the measurement data of each acceleration sensor 203 (e.g., the measurement signal measured by the acceleration sensor 203, frequency spectrum data obtained by Fourier transforming the measurement signal, etc.), peak data of a specific frequency range obtained using the measurement signal (e.g., data including frequency, value indicating peak height, date and time information, group number, etc.), etc.

The peak data analysis unit 303 performs a Fourier transform on the measurement signal measured by the acceleration sensor 203 and stored in the data storage unit 306 to obtain a frequency spectrum. The peak data analysis unit 303 also uses the concept of peak height to obtain a frequency that is a maximum value (hereinafter referred to as a "peak frequency") from among the many peaks that appear in the frequency spectrum. The peak data analysis unit 303 then sorts the peak data in order of peak height and removes unnecessary peaks, extracting the top peak data and storing it in the data storage unit 306. A detailed explanation of the peak data analysis unit 303 will be given in the section on operation. When the peak data analysis unit 303 creates the peak data, the group number does not need to be determined. The date and time information is information that indicates the date and time of measurement by the acceleration sensor 203.

The grouping processing unit 304 also reads out multiple peak data measured in the past from the data storage unit 306, assigns a group number to each specific frequency region, and consolidates (groups) the peak data for each specific frequency region. The grouping processing unit 304 also repeatedly performs grouping using a large amount of peak data from specific frequency regions, and determines the peak characteristics of a specific frequency region that is estimated to be the natural frequency of the structure. A detailed explanation of the grouping process will be given in the section on operation.

The natural frequency group extraction unit 307 extracts frequencies that are estimated to belong to a natural frequency group from the multiple frequencies obtained by grouping. A detailed explanation of the method for extracting a natural frequency group will be given in the section on operation.

Here, "natural frequency group" refers to multiple frequencies that are considered to be natural frequencies or integer multiples of natural frequencies among multiple specific frequency ranges.

The natural frequency estimation unit 308 estimates the natural frequency based on multiple frequencies belonging to the natural frequency group extracted by the natural frequency group extraction unit 307.

For example, the natural frequency estimation unit 308 uses the characteristic that peaks appear at frequencies that are integer multiples of the natural frequency to find the frequency difference between each group in the natural frequency group. In other words, it finds the frequency difference between a certain group and an adjacent group. The natural frequency estimation unit 308 then finds the average value of the frequency difference values between the groups (referred to as the average frequency difference value between groups), and when the average frequency difference value between the groups is a certain number (or a value that is close to a certain number), it estimates that certain number as the natural frequency. An example of a method for estimating the natural frequency will be described in detail in the section on operation.

(A-1-3) Internal Configuration of Sensor Device FIG. 3 is a diagram showing the internal configuration of the sensor device 20 according to the embodiment.

In FIG. 3, the sensor device 20 has a communication unit 201, a control unit 202, an acceleration sensor 203, a schedule determination unit 204, a timer unit 205, a clock unit 206, and a data storage unit 207.

The sensor device 20 may be configured as hardware, and some of the components of the sensor device 20 may be configured as software.

The communication unit 201 is a communication interface that communicates with the network connected to the parent unit 10.

The acceleration sensor 203 measures vibrations of a structure such as a cable or a bridge. The acceleration sensor 203 measures instantaneous acceleration and provides a measurement signal to the control unit 202. One sensor device 20 may be equipped with multiple acceleration sensors 203. Note that the sensor types are not limited to this. For example, the sensor device 20 may be equipped with various types of sensors, such as a temperature sensor, a humidity sensor, a vibration sensor, an infrared image sensor, etc., in addition to or instead of the acceleration sensor 203.

The control unit 202 is a device and processing unit that controls various functions in the sensor device 20. The control unit 202 has, for example, a CPU, ROM, RAM, EEPROM, an input/output interface, etc. The CPU may execute a processing program (e.g., a measurement program, etc.) stored in the ROM to realize processing.

The control unit 202 also receives information about the measurement operation from the monitoring device 30 via the communication unit 201, and provides measurement operation schedule information for operating the acceleration sensor 203 to the schedule determination unit 204.

The schedule determination unit 204 determines the periodic measurement timing for operating the acceleration sensor 203 based on the measurement operation schedule information received from the monitoring device 30. The schedule determination unit 204 sets the periodic measurement timing of the acceleration sensor 203 in the timer unit 205.

The timer unit 205 manages the measurement timing set by the schedule determination unit 204. The timer unit 205 monitors the current time of the clock unit 206, and when the time reaches the measurement time of the measurement timing, the timer unit 205 activates the control unit 202. This activates the control unit 202, which then activates the acceleration sensor 203, which then operates. Note that at times other than the measurement time, the sensor device 20 is in a sleep state, and all components other than the timer unit 205 and clock unit 206 are put into a sleep state. This makes it possible to reduce power consumption.

The clock unit 206 is a clock that provides the current time.

(A-2) Operation of the Embodiment First, the overall processing in the monitoring system 1 will be briefly described.

The sensor device 20 periodically determines the measurement timing for operating the acceleration sensor 203 based on the measurement operation schedule. In the sensor device 20, when the measurement timing time arrives, the acceleration sensor 203 operates for a predetermined time. In the operating state, the acceleration sensor 203 measures data indicating the characteristics of the structure, such as acceleration due to vibration. The communication unit 201 transmits information including the measurement signal (measurement value) measured by the acceleration sensor 203 to the parent unit 10, and the parent unit 10 transfers the information received from the sensor device 20 to the monitoring device 30.

Next, an example of the operation of the monitoring method in the monitoring device 30 according to the embodiment will be described with reference to the drawings.

Figure 4 is a flowchart showing the processing operation of the monitoring device 30 according to the embodiment. Below, the processing in the monitoring device 30 will be explained based on the procedure shown in Figure 4.

[S1: Derivation of peak height]
The monitoring device 30 receives a measurement signal (measurement value) measured by the acceleration sensor 203 of the sensor device 20. In the monitoring device 30, the peak data analysis unit 303 performs a Fourier transform on the measurement signal (time-series vibration waveform) to obtain a frequency spectrum. The monitoring device 30 analyzes the frequency spectrum to identify the natural frequency of the structure, and detects a change in the strength (amplitude) of the signal component of the natural frequency to determine an abnormality in the structure.

For example, for each measurement by the acceleration sensor 203, the monitoring device 30 extracts peaks from the frequency spectrum obtained by Fourier transforming the measurement signal, and obtains peak data for each specific frequency range.

The frequency spectrum has many peaks. The natural frequency is generally present among multiple peak frequencies that appear as maximum values in the frequency spectrum. However, even among maximum values in the frequency spectrum, there are some inappropriate maximum values that are not suitable as peaks to use in determining abnormalities in a structure.

For example, due to the characteristics of the Fourier transform, many small peaks may exist at frequencies before and after a large peak. These small peaks will disappear if the amount of data used in the Fourier transform is increased drastically, and it is therefore inappropriate to use them as peaks.

Therefore, in this embodiment, the peak data analysis unit 303 uses the concept of "peak height" to derive the "peak height" of each of the many peaks present in the frequency spectrum.

Here, one example of deriving the "peak height" is the concept of "prominence." The process of deriving prominence will be explained using Figures 5 and 6.

Figure 5 (A) is an example of a frequency spectrum. For ease of explanation, the left end point of the frequency spectrum is designated as "a" and its right end point as "g". The five maximum values from "left end a" to "right end g" are designated as peak numbers "1" to "5", and the five minimum values are designated as "b" to "f".

6, the peak data analysis unit 303 derives the prominence indicating the height of each peak. First, the peak data analysis unit 303 selects a peak (S101), moves the peak to the left and right until the value reaches the next state, and derives the end points of the horizontal line (S102).
- The value crosses a higher peak - The value reaches the left or right edge.

Next, the peak data analysis unit 303 determines the minimum points (minimum values) in the left and right sections of the peak derived in S102 (S103), and sets the higher of the two minimum values derived in S103 as the reference level (maximum/minimum point) (S104). The peak data analysis unit 303 sets the difference between the reference level and the peak as the height (prominence) of the peak (S105).

The peak data analysis unit 303 determines whether or not all peaks have been selected (S106), and if all peaks have not been selected (S106/No), the peak data analysis unit 303 proceeds to S101 and repeats the process. On the other hand, if all peaks have been selected (S106/Yes), the peak data analysis unit 303 ends the process.

For example, we will explain an example of a method for deriving the prominence of the peak with "Peak Number: 1" in Figure 5 (A).

First, the peak with "peak number: 1" is selected (S101). When the value is moved to the left from this peak, the value reaches "left end a", so "left end a" becomes the end point of the horizontal line of the peak left section. On the other hand, when the value is moved to the right from the peak, it intersects with the peak with "peak number: 2", which is higher than the peak with "peak number: 1", so the point where the value intersects with peak 2 becomes the end point of the horizontal line of the peak right section (S102).

Next, since the minimum point of the left section of the peak is "left end a" and the minimum point of the right section of the peak is "b" (S103), the higher of "left end a" and "b" is set as the reference level (S104). Then, the height "P1" from the reference level "b" to "peak number: 1" is set as the height (prominence) of the peak of "peak number: 1" (S105). The peak data analysis unit 303 performs the above-mentioned processing on all peaks, and the results are shown in Figure 5 (B).

[S2: Removal of unnecessary peaks]
Next, the peak data analysis unit 303 compares the height of each peak in the frequency spectrum with the height of the peaks that appear before and after each peak, and, taking into account the difference in frequency, deletes the smaller peak if the ratio is below a certain level.

For example, in the case of Figure 5(A), the height of the peak with "peak number: 3" is smaller than the height of the peak with the adjacent "peak number: 4", so the peak with "peak number: 3" is subject to deletion.

More specifically, the peak data analysis unit 303 calculates the ratio between the peak height P3 of "peak number: 3" and the peak height P4 of "peak number: 4", and when this ratio is equal to or less than a threshold, it determines that the peak of "peak number: 3" has a small height and is to be deleted. As another method, the peak data analysis unit 303 may calculate the difference between the peak height P3 of "peak number: 3" and the peak height P4 of "peak number: 4", and when this difference value is equal to or less than a threshold, it may determine that the peak of "peak number: 3" has a small height. Note that the threshold used when determining which peaks to adopt based on their peak heights may be changeable as appropriate.

As another method, the peak data analysis unit 303 may sort the peak data in order of peak height and use only a fixed proportion of the data from the top as peaks.

Figure 7 is an example of a frequency spectrum. The horizontal axis is frequency (Hz), and the vertical axis is amplitude spectrum (gal x s). For example, high peaks appear near 8 Hz and 11 Hz. In contrast, for example, peaks appear near 7 Hz and 10 Hz, but the height of the peak near 7 Hz is lower than that of the peak near 8 Hz, and similarly, the height of the peak near 10 Hz is lower than that of the peak near 11 Hz, so the peaks near 7 Hz and 11 Hz are not adopted as peaks in this embodiment. In other words, the peak data analysis unit 303 deletes the peaks near 7 Hz and 11 Hz.

[S3: Creation of peak data]
The peak data analysis unit 303 creates peak data for the remaining multiple peaks by deleting unnecessary peaks from the peaks present in the frequency spectrum. For example, the peak data analysis unit 303 extracts peaks from a frequency spectrum based on measurements taken in one day, and creates peak data for each peak. The peak data analysis unit 303 then stores the peak data for each peak in the data storage unit 306.

The peak data here is information for each peak, and includes the frequency at which the peak appears (peak frequency), the peak height value, date information, group number, etc. Note that at this stage, the group number for each peak has not been determined, and no group number has been assigned.

[S4: Grouping process]
The grouping processing unit 304 reads out the peak data for multiple days from the data storage unit 306 and sorts all the peak data in order of frequency. The grouping processing unit 304 then assigns a group number to all the peak data. For example, in the initial state, a unique group number is assigned in order to each of the peak data sorted in order of frequency.

Next, the grouping processing unit 304 focuses on the two groups with the smallest frequency difference among the peak data with group numbers. The grouping processing unit 304 then examines all peak data in the two groups and determines that the two groups do not contain data for the same day. If there is no data for the same day, the grouping processing unit 304 combines the two groups. Note that if there is data for the same day, the groups are not combined.

The grouping processing unit 304 performs the above-mentioned group combining process for all groups. The combined group is also considered to be a new group, and grouping is repeated by determining whether there is data for the same day between the combined group and other groups. By repeatedly performing the grouping process, peak data with close frequencies are placed in the same group in order, and frequency bands are naturally formed.

The grouping processing unit 304 compares the peak data of two groups with adjacent group numbers (or frequencies) even for groups formed as one group, and if there is no data for the same day, combines the groups.

For example, being in different groups means that peak data for the same day exists in both groups for at least one day. Here, the existence of peak data for the same day in both groups is referred to as "overlap." The grouping processing unit 304 checks the proportion of "overlap" that exists in the group (for example, the number of overlaps relative to the total number). The grouping processing unit 304 also compares the number of peak data belonging to the two groups.

For example, the grouping processing unit 304 compares the number of peak data belonging to each of the two groups. For example, assume that one group has a large number of peak data and the other group has a small number of peak data. In this case, if the number of overlaps in the two groups is also small (if the number of overlaps is equal to or less than a threshold), the grouping processing unit 304 combines the two groups.

For example, if the "peak height" of the peak data overlapping two groups is clearly low (if the peak height is below a threshold value), the grouping processing unit 304 considers it to be an unimportant peak and combines the two groups.

In this way, the grouping processing unit 304 takes into account peak data that overlaps in two groups, even if the groups are formed as one group.

[S5: Group Selection]
In many cases, the above-described grouping process results in multiple groups. Next, the grouping processor 304 selects a group to be used as the natural frequency of the structure.

An example of a method for selecting a group by the grouping processing unit 304 will be described. The method for selecting a group is not limited to using one criterion (or method), but may use multiple criteria (or methods). For example, the grouping processing unit 304 selects a group in which a certain percentage or more of peak data exists relative to the number of measurements. In another example, the grouping processing unit 304 selects a group to which peak data with a peak height equal to or greater than a threshold value belongs.

The "certain percentage or more" rule is particularly important when monitoring infrastructure structures that have no vibration source. For example, if there is no vehicle traffic due to a road closure or if there is no wind, a peak may not appear even if there is nothing abnormal with a bridge or other structure. Therefore, the absence of a peak cannot be concluded as necessarily indicating an abnormality.

Therefore, the "certain percentage or more" is set in order to select not only frequencies (groups) with many prominent or frequent peaks, but also frequencies (groups) where no peaks appear under normal operating conditions of the structure.

Of course, the frequencies of prominent peaks and peaks that appear frequently are also important frequencies (groups), so these groups must also be selected. The value of the fixed percentage is not particularly limited, and may be determined appropriately depending on the condition and environment of the structure to be monitored.

Figure 8 shows an example of data results measured daily by the monitoring device 30. The horizontal axis represents the date, and the vertical axis represents the frequency (Hz). In Figure 8, the size of the circle indicates the "peak height," with a larger diameter of the circle indicating a higher "peak height."

Figure 9 shows an example of peak data belonging to some groups from the data results in Figure 8. For example, the "f" column shows the frequency, the "p" column shows the peak height, and the numbers "f" and "p" show the group number. For example, in Figure 8, around 6 Hz is group 9, around 8 Hz is group 10, and around 11 Hz is group 11.

For example, group 11 around 11 Hz is a group containing peak data with peak heights equal to or greater than a threshold. For example, group 10 around 8 Hz is a group containing a certain percentage or more of peak data relative to the number of measurements. In this way, it is possible to select groups in which no peaks normally appear (or where peaks are unlikely to appear).

[S6: Extraction of Eigenfrequency Group]
The natural frequency group extracting unit 307 extracts a natural frequency group from the plurality of frequencies obtained by grouping.

Various methods can be applied to extract the natural frequency groups. In principle, the extraction of the natural frequency groups is premised on the fact that peaks appear at integer multiples of the natural frequency due to vibration of the structure. Ideally, peaks should always appear at integer multiples of the natural frequency, but there are other factors that may prevent the peaks from appearing at equal intervals.

For example, the following formula (1) is formula 1 described in Non-Patent Document 1, but this example also shows that ^fi2 contains a component proportional to ⁱ⁴ (i.e., a component in which the square of the natural frequency is proportional to the fourth power of the integer value (order value)). However, in many cases, the influence of these terms is considered to be small. When the number of parameters increases in this way, a configuration can be adopted in which the most appropriate parameters are estimated using the distribution of peaks obtained from multiple spectra. In addition, a configuration can be adopted in which multiple parameters including the natural frequency are simultaneously estimated using a method such as the least squares method.

式（１）において、ｆｉは、固有振動数［Ｈｚ］、iは次数、Ｔは張力［Ｎ］、ＥＩは曲げ剛性［Ｎ・ｍ^２］、Ｌはケーブル長［ｍ］、ρＡは単位長さ当たりの重量［ｋｇ／ｍ］である。

In equation (1), fi is the natural frequency [Hz], i is the order, T is the tension [N], EI is the bending rigidity [N·m ² ], L is the cable length [m], and ρA is the weight per unit length [kg/m].

[S7: Estimation of natural frequency]
First, the natural frequency group extracting unit 307 obtains the average frequency of a plurality of grouped specific frequency regions (frequency bands). That is, the natural frequency group extracting unit 307 obtains the average value of the frequencies of each group.

Next, the eigenfrequency group extraction unit 307 calculates the difference in average frequency between a group and an adjacent group.

The natural frequency estimation unit 308 averages the difference values of the average frequencies between the groups obtained by the natural frequency group extraction unit 307 to obtain an average difference value. In other words, the natural frequency estimation unit 308 obtains the interval between the frequencies at which peaks appear (hereinafter, also referred to as the "estimated interval").

The natural frequency estimation unit 308 obtains the distribution of the average difference values between each group, and assumes that the most frequently occurring average difference value is the natural frequency. By utilizing the characteristic that a peak appears at a frequency that is an integer multiple of the natural frequency, if the groups are spaced at approximately equal intervals, the intervals can be considered to be the natural frequency. Therefore, here, the most frequently occurring average difference value is derived from the average difference values between the groups, and this average difference value is assumed to be the natural frequency.

The natural frequency estimation unit 308 then divides the average frequency of each group by the assumed natural frequency and verifies that the quotient is close to an integer value. It is important to note that the average frequencies are not necessarily equally spaced, since peaks may not be found at certain multiples.

If it cannot be confirmed that the quotient is close to an integer value, the assumed natural frequency is incorrect, so the natural frequency estimation unit 308 eliminates the abnormal data and starts over from the natural frequency assumption.

If it is confirmed that the quotient is close to an integer value, the natural frequency estimation unit 308 eliminates outliers (i.e. peaks that are not part of the natural frequency group) within the group and determines the natural frequency.

Here, we explain how to check whether the quotient is close to an integer value. For example, it is often the case that a sensor cannot measure the natural frequency due to sensor characteristics such as being unable to measure frequency components below a certain frequency. Even in such cases, if the intervals between the frequencies at which peaks appear are constant, the natural frequency can be estimated.

For example, suppose the natural frequency is 0.8 [Hz]. In this case, peaks only appear at 3.2 [Hz], 4 [Hz], and 4.8 [Hz]. In such a case, the difference value of the frequencies at which each peak appears is 0.8 [Hz], so the natural frequency estimation unit 308 divides the frequencies 3.2 [Hz], 4 [Hz], and 4.8 [Hz] by the difference value 0.8 [Hz] to obtain integer multiples of 4 (times), 5 (times), and 6 (times). In this case, the natural frequency can be determined to be 0.8 Hz.

However, in reality, the frequencies at which peaks appear are often shifted, such as 3.3 Hz, 4.1 Hz, 4.7 Hz, etc., so a clean integer value (multiple value) cannot be obtained, making it difficult to mechanically estimate the natural frequency.

Therefore, for example, it is possible to confirm that the quotient is close to an integer value (multiple value) by using a predetermined rounding method (e.g., rounding up, rounding down, rounding up, etc.). In this case, in order to consider the quotient to be close to the expected integer value (multiple value), a certain range may be set in advance, and the quotient may be recognized as an integer value if it is within that range.

As a different configuration, an approximate natural frequency can be specified in advance, and a group having an average frequency close to a natural frequency that is an integer multiple of the specified value can be adopted as the natural frequency group.

[S8: Estimation of tension]
The control unit 302 estimates physical values such as tension using the natural frequency.

This makes it possible to estimate the tension acting on the cable from the value of the natural frequency, for example, and to determine whether the tension is as designed. Note that in some cases, a change in the value of the natural frequency itself may indicate that there is an abnormality in the cable.

In addition, in the embodiment, an example is given of determining the tension value of a cable, but the value is not limited to tension as long as a physical value for determining the soundness of a structure can be derived based on the natural frequency.

For example, as exemplified by the above-mentioned formula (1), it is known that there is a certain relationship between the natural frequency and the tension acting on a cable. When the structure to be judged for anomalies is a cable, it may be necessary to consider factors other than the natural frequency, and it is desirable to derive the tension while including the other factors and judge the anomaly. For example, in formula (1) which includes other factors, the tension can be derived using parameters that do not change, such as the cable length.

For example, the control unit 302 may estimate parameters from the distribution of the natural frequency group using a method such as the least squares method, and use the estimated parameters to determine tension, etc.

As a modified example, the control unit 302 may have a function to issue a warning if the tension changes beyond a predetermined range, thereby reducing the monitoring work. In this case, the estimation of the tension contains a certain amount of error. Therefore, rather than issuing a warning based on the result of only one measurement, when a change in tension is detected, the acceleration sensor 203 measures again to confirm that the abnormality (change in tension) continues. Then, if the change in tension continues, some measure is required, such as having the control unit 302 issue a warning.

In addition, when multiple acceleration sensors 203 are attached to a structure such as a cable, it is also possible to provide a function for making a judgment by combining the estimation results of other sensors. For example, two acceleration sensors 203 are attached to a cable, and the measurement values of each acceleration sensor 203 are used to estimate the natural frequency and determine the tension. In this case, if both tensions estimated using the measurement values of each acceleration sensor 203 have changed beyond the range, it may be determined that an abnormality has occurred in the cable.

In addition, by estimating the natural frequency using a large amount of measurement data obtained by repeating measurements by the acceleration sensor 203 over a long period of time, the natural frequency can be estimated with high accuracy.

[Example]
FIG. 10 is a diagram showing frequencies for each group in the embodiment.

In Figure 10, for example, the 0th group is around 0.3 Hz, and the frequency where a peak appears is shown in the "f0" column. Figure 10 shows the frequencies in the group f0 to f7.

First, the natural frequency group extraction unit 307 excludes f1 and f4, which have a small amount of data, from among the groups f0 to f7 in FIG. 10, and selects the six groups f0, f2 to f3, and f5 to f7 as candidates for natural frequency groups.

For example, in this example, a case is illustrated in which groups (f1, f4) with a data count less than a threshold (e.g., 6) are excluded, but the decision to extract natural frequency group candidates is not limited to a decision based on the data count. For example, the decision may be based on the ratio of the number of data when a peak is detected to the total number of samples (i.e., a comparison between a value indicating the peak detection ratio and a threshold). In addition, since selecting natural frequency group candidates is very difficult, it is desirable to be able to set the threshold as appropriate according to operation.

Next, the eigenfrequency group extraction unit 307 calculates the average frequency of each group of the eigenfrequency group candidates for each group. This corresponds to the value shown in the "(A) Average frequency value" column in FIG. 10.

Furthermore, the eigenfrequency group extraction unit 307 finds the difference in average frequency between a certain group and an adjacent group. This corresponds to the value shown in the "(B) Difference value" column in FIG. 10. Note that since there is no group before the group f0, the "(B) Difference value" column for the group f0 shows the average frequency value.

Next, the eigenfrequency group extraction unit 307 calculates the average value of the difference values of each group. In this example, the average difference value D obtained by averaging the difference values of each group is 0.757558 [Hz].

Here, there are values such as "1.381701" and "1.217227" which greatly exceed the average difference value D (0.75755 Hz). The natural frequency estimation unit 308 divides these values by the average difference value, and the quotient obtained by this is rounded (for example, rounded up, rounded down, etc.) to obtain a numerical value of "2".

For example, if you divide "1.381701" by the average difference value D "0.75755", the quotient is approximately 1.82391, which becomes "2" when rounded off to the first decimal place.

In that case, we can infer that the difference between f0 and f2, "1.381701," means that one peak was not measured. The same is true for "1.217227."

Therefore, the natural frequency estimation unit 308 divides each of "1.381701" and "1.217227" by "2" to obtain the difference value of the average frequency of each group again. This corresponds to the value shown in the "(C) Estimation interval" column in Figure 10.

Then, the natural frequency estimation unit 308 calculates the average value of the "(C) Estimation Interval" for each group, which is "0.54098". This value becomes the natural frequency estimate C.

In the above example, the value exceeding the average difference value D is divided by the average difference value D to obtain a numerical value of "2", but it may also be rounded down to an integer value such as 3 or 4. For example, if the integer value obtained by rounding is "3", it can be determined that two peaks have not been measured between the difference values. In this case, the difference value can be divided by "3", and the average value of the difference values for each group can be used as the estimate of the natural frequency.

Furthermore, the natural frequency estimation unit 308 determines for each group how many times the average frequency value (A) of each group is larger than the estimated natural frequency value C. This corresponds to the value in the "(D) How many times larger than C" column in Figure 10.

For example, the value obtained by dividing the average frequency value of the f0 group, "0.304026", by the estimated natural frequency C, "0.54098", is "0.561991". The other groups are calculated in the same way.

The natural frequency estimation unit 308 performs rounding, etc., and if the value in the "(D) How many times C" column in FIG. 10 is close to an integer value, it can be determined that the estimated natural frequency value is correct. For example, to determine whether it is an integer value, the difference between the expected integer value and the actual value can be taken, and a threshold value or the like can be used to determine how far it deviates. Also, rounding can be used as appropriate, for example by rounding to two decimal places.

In this example, the values f0, f5, f6, and f7 are not close to integer values. This means that the estimation of the natural frequency has failed.

Therefore, the natural frequency group extraction unit 307 and the natural frequency estimation unit 308 redo the removal of the group that is thought to be the cause of the failure in estimating the natural frequency (extraction of the natural frequency group).

Various methods can be applied to redo the estimation of the natural frequency depending on the situation, but for example, in the above example, f0, which deviates the most from the expected integer value, is excluded and the natural frequency estimate is calculated again.

In other words, the natural frequency estimation unit 308 again calculates the difference between each group ("(C2) Estimation interval" in FIG. 10) to calculate the natural frequency estimate C2. Here, since this is the second estimation of the natural frequency, it is expressed as estimate C2, and the natural frequency estimate C2 is "0.562583".

Then, the natural frequency estimation unit 308 again determines for each group how many times the average frequency value (A) of each group is the natural frequency estimate C2. For example, the values in the "(D2) How many times C2" column in FIG. 10 are roughly integer values, and f2 can be considered to be 3 times, f3 4 times, f5 6 times, f6 7 times, and f7 8 times. In this example, the natural frequency is the estimate C2 "0.562583", and f2, f3, f5, f6, and f7 are determined as the natural frequency group.

In the above example, the natural frequency estimation unit 308 sets the estimated value C2 "0.562583" as the natural frequency, but this is not limiting. The natural frequency estimation unit 308 may set a new value based on the average difference value between each group belonging to the determined natural frequency group as the natural frequency. This makes it possible to determine a more accurate natural frequency.

The control unit 302 then applies the method described above to estimate the tension on the cable based on the obtained natural frequency.

(A-3) Effects of the embodiment As described above, according to this embodiment, the natural frequency can be estimated from the frequency spectrum obtained by Fourier transforming the measurement value of the sensor, and physical values such as tension can be estimated based on the natural frequency, making it possible to automatically perform abnormality judgment on the structure.

In addition, according to the embodiment, the estimated tension and other physical values are used to determine abnormalities in the structure, and the results are used to automatically issue warnings, thereby reducing the amount of monitoring work.

(B) Other Embodiments Although various modified embodiments have been mentioned in the above-described embodiment, the present invention can also be applied to the following modified embodiments.

(B-1) Figures 8 and 9 show data obtained from one axis of a single sensor, but acceleration sensors can generally obtain data on three axes, x, y, and z, simultaneously. The direction of the three axes in which the natural vibration of a structure appears depends on the installation conditions. Also, not all peaks will necessarily appear in a single direction. For example, there may be a situation in which a peak four times the natural frequency appears on the X axis and five times the natural frequency appears on the Y axis. Therefore, a configuration can be used in which data from multiple axes and sensors is integrated.

(B-2) In the above embodiment, first, the natural frequency group extraction unit 307 extracts candidates for the natural frequency group, and then the natural frequency estimation unit 308 estimates the natural frequency. However, the extraction of the natural frequency group candidates may be omitted. For example, without extracting the natural frequency group candidates, the natural frequency estimation unit 308 may find the average frequency value of each group, and estimate an integer multiple value in a similar manner to find the natural frequency. In this case, there is a process of repeatedly redoing the estimation of the natural frequency as necessary, and groups that are deemed unnecessary at that time may be excluded.

(B-3) In the above embodiment, the monitoring target is one and the same structure, but the present invention can also be used as is when monitoring multiple structures constructed at the same time as one group.

(B-4) The various functions of the monitoring device described above may be performed by the same physical device, or may be distributed and processed by separate devices.

1... monitoring system, 10... parent unit, 20... sensor equipment, 30... monitoring device, 201... communication unit, 202... control unit, 203... acceleration sensor, 204... schedule determination unit, 205... timer unit, 206... clock unit, 207... data storage unit, 301... communication unit, 302... control unit, 303... peak analysis unit, 304... grouping processing unit, 306... data accumulation unit, 307... natural frequency group extraction unit, 308... natural frequency estimation unit.

Claims (10)

an analysis means for detecting peaks appearing in a frequency spectrum obtained based on a measurement signal from a sensor installed in the structure, and obtaining information about each peak;
A storage means for storing information about each of the peaks for each measurement;
a grouping means for grouping the frequencies at which the peaks are detected by frequency bands based on information about the peaks obtained through a plurality of measurements;
a natural frequency group extracting means for extracting a plurality of frequency bands belonging to a natural frequency group from the plurality of frequency bands grouped by the grouping means;
and a natural frequency estimation means for estimating a natural frequency based on a plurality of frequency bands belonging to the extracted natural frequency group. The monitoring device according to claim 1, characterized in that the natural frequency estimation means estimates the natural frequency based on the estimation interval between adjacent peak frequencies in the plurality of frequency bands belonging to the natural frequency group. the natural frequency group extraction means calculates a difference value between adjacent frequency bands using an average frequency value of each of the grouped frequency bands;
3. The monitoring device according to claim 1, wherein the natural frequency estimation means estimates the natural frequency based on an average difference value obtained by averaging the difference values between the frequency bands adjacent to each other. The natural frequency estimation means
When a value obtained by dividing the difference value between the adjacent frequency bands by the average difference value for each of the grouped frequency bands is close to a value corresponding to an integer ,
The difference value is divided by the integer that approximates the difference value to obtain a value of the estimation interval, and the estimation interval average value obtained by averaging the values of the estimation interval for each of the grouped frequency bands is estimated as the characteristic frequency.
4. The monitoring device according to claim 3. The natural frequency estimation means
When a value obtained by dividing the difference value between the adjacent frequency bands by the average difference value for each of the grouped frequency bands is not approximate to a value corresponding to an integer ,
One of the grouped frequency bands is excluded, and the natural frequency is estimated again.
5. The monitoring device according to claim 3 or 4. A monitoring device according to any one of claims 1 to 5, characterized in that it is provided with a control means for deriving a numerical value indicating the state of the structure based on the natural frequency. The monitoring device according to claim 6, characterized in that the control means estimates parameters related to the derivation of a numerical value indicating the state of the structure based on the distribution of the natural frequency group, and derives the numerical value indicating the state of the structure using the parameters. The monitoring device according to claim 6 or 7, characterized in that the control means issues a warning when the derived numerical value indicating the state of the structure deviates from a predetermined range or deviates from the range of values obtained in past measurements. Computer,
an analysis means for detecting peaks appearing in a frequency spectrum obtained based on a measurement signal from a sensor installed in the structure, and obtaining information about each peak;
A storage means for storing information about each of the peaks for each measurement;
a grouping means for grouping the frequencies at which the peaks are detected by frequency bands based on information about the peaks obtained through a plurality of measurements;
a natural frequency group extracting means for extracting a plurality of frequency bands belonging to a natural frequency group from the plurality of frequency bands grouped by the grouping means;
a monitoring program causing the computer to function as a natural frequency estimation means for estimating a natural frequency based on a plurality of frequency bands belonging to the extracted natural frequency group. The analysis means detects peaks appearing in a frequency spectrum obtained based on a measurement signal from a sensor provided in the structure, and obtains information about each peak;
A storage means stores information about each of the peaks for each measurement,
a grouping means for grouping the frequencies at which the peaks are detected by frequency bands based on information about the peaks obtained over a plurality of measurements;
a natural frequency group extracting means for extracting a plurality of frequency bands belonging to the natural frequency group from the plurality of frequency bands grouped by the grouping means;
a natural frequency estimating means for estimating a natural frequency based on a plurality of frequency bands belonging to the extracted natural frequency group.

Images